Section 1 made the case for why sleep matters. This section takes the science deeper, organised around six territories: the architecture of sleep itself, how much sleep is actually needed, the difference between time in bed and time well-slept, the body’s internal clock, the consistency of sleep timing, and the environment that shapes whether sleep happens well or badly. Each of these maps onto a lever members can pull, and each comes with its own contested edges.
The first thing to understand about sleep is that it is not a single state. Across a normal night, the brain moves through a sequence of distinct phases, looping through what sleep scientists call cycles. A typical adult cycle lasts around 90 minutes and includes light sleep (stages N1 and N2), deep sleep (stage N3, also called slow-wave sleep), and a phase called REM (rapid eye movement) sleep, the phase in which most dreaming occurs. Most adults move through four to five of these cycles per night.
The cycles change shape as the night progresses. The first half of the night is dominated by deep sleep — the long, slow brain waves that give slow-wave sleep its name. This is the most physically restorative phase: the body releases growth hormone, repairs tissue, consolidates motor learning, and the brain’s waste-clearance system runs at full capacity. The second half of the night is dominated by REM sleep, where dreams are most vivid and where most emotional processing and declarative memory consolidation happens. Lose the start of the night and you lose deep sleep; lose the tail end and you lose REM. Both losses matter, and they cost different things.
The brain’s waste-clearance system is called the glymphatic system, and its discovery is one of the most important findings in sleep research of the past fifteen years. During deep sleep, the spaces between brain cells expand by roughly 60 per cent, allowing cerebrospinal fluid to flush metabolic waste — including the beta-amyloid proteins implicated in Alzheimer’s disease — out of the brain at rates several times higher than during waking hours. The original mouse study, published in Science in 2013 by a team at the University of Rochester, found that glymphatic clearance of beta-amyloid was approximately twice as fast during sleep as during waking. Subsequent work has extended and refined the finding in humans.
The implication is that deep sleep is not a passive recovery phase. It is when the brain takes the rubbish out. Skip enough deep sleep over enough years and the rubbish accumulates faster than it can be cleared. This is part of why the dementia evidence on chronic sleep loss is as strong as it is.
Sleep architecture also changes with age in predictable ways. Total sleep time drops modestly, by around half an hour to an hour by the seventh decade. More consequentially, the proportion of deep sleep declines — sometimes by as much as 60 to 70 per cent between the third and seventh decades. Older adults spend more of the night in lighter sleep stages, wake more often during the night, and produce less of the deep slow-wave sleep that does the heavy biological work. Some of this is genuinely normal age-related change. But the consequences for waste clearance are serious enough that preserving as much deep sleep as possible into older age is one of the most defensible single goals of a longevity-focused sleep practice.
The implication is that deep sleep is not a passive recovery phase. It is when the brain takes the rubbish out. Skip enough deep sleep over enough years and the rubbish accumulates faster than it can be cleared. This is part of why the dementia evidence on chronic sleep loss is as strong as it is.
Sleep architecture also changes with age in predictable ways. Total sleep time drops modestly, by around half an hour to an hour by the seventh decade. More consequentially, the proportion of deep sleep declines — sometimes by as much as 60 to 70 per cent between the third and seventh decades. Older adults spend more of the night in lighter sleep stages, wake more often during the night, and produce less of the deep slow-wave sleep that does the heavy biological work. Some of this is genuinely normal age-related change. But the consequences for waste clearance are serious enough that preserving as much deep sleep as possible into older age is one of the most defensible single goals of a longevity-focused sleep practice.
How much sleep an adult needs is the question most asked about sleep, and it has a reasonably clear answer: around seven to nine hours for most adults, with seven hours being the point of lowest mortality risk in the large epidemiological studies, and very few adults genuinely thriving on less than six.
The foundational evidence is a 2010 meta-analysis led by Francesco Cappuccio at the University of Warwick, pooling 16 prospective cohort studies covering more than 1.3 million participants. Short sleep increased all-cause mortality risk by 12 per cent (relative risk 1.12); long sleep increased it by 30 per cent (RR 1.30). The relationship was U-shaped, with the lowest risk centred around seven hours.
Subsequent dose-response analyses have refined the picture with finer granularity. A 2016 meta-analysis by Shen and colleagues, pooling 35 studies and over 1.5 million participants, quantified the risk at specific sleep durations: compared with seven hours per night, those sleeping six hours had a 1 per cent higher mortality risk, five hours a 4 per cent higher risk, eight hours a 7 per cent higher risk, nine hours a 21 per cent higher risk, and ten hours a 37 per cent higher risk. A 2017 dose-response meta-analysis published in the Journal of the American Heart Association, covering 74 cohort studies, confirmed this pattern across all-cause mortality, cardiovascular disease, coronary heart disease and stroke, finding that each hour of sleep below seven added roughly 6 per cent to mortality risk, while each hour above seven added roughly 13 per cent. A separate 2017 analysis using flexible non-linear modelling reached the same conclusion. A 2025 meta-analysis covering 79 studies, the most recent large synthesis, found short sleep associated with a 14 per cent mortality increase and long sleep with a 34 per cent increase — with the long-sleep effect notably larger in women than men.
Two patterns in this data are worth noticing. The first is that the curve is asymmetric: the long-sleep tail is consistently steeper than the short-sleep tail. The second is that the long-sleep association is almost certainly not all causal. Long habitual sleep is often a marker for undiagnosed illness, depression, or sedentary lifestyle, which is why most sleep researchers are more cautious about recommending against long sleep than about recommending against short sleep. Sleeping nine hours when you feel unwell is a symptom, not a lifestyle choice. What the short-sleep end of the curve tells us is more actionable.
Whether long habitual sleep causes increased mortality, or whether it is simply a marker for other health problems that cause both long sleep and higher mortality, remains genuinely disputed. Three of the four large meta-analyses cited in this section flag reverse causation as a likely contributor to the long-sleep tail. The short-sleep half of the U-curve has stronger causal evidence (experimental sleep restriction produces the hypothesised mechanisms — metabolic, cardiovascular, cognitive — within days to weeks). The long-sleep half does not have comparable experimental evidence. Forever Well’s view: focus on not under-sleeping. If you habitually need more than nine hours, treat that as a signal to investigate, not as a problem to be solved by sleeping less.
Dedicated meta-analyses of the short-sleep half confirm that habitual short sleep is associated with increased risk of diabetes, hypertension, cardiovascular disease, coronary heart disease, and obesity, in addition to mortality. Dedicated meta-analyses of the long-sleep half find similar associations, though weaker for hypertension.
The dementia evidence reaches the same conclusion from a different angle. The 2021 Whitehall II study, following nearly 8,000 British civil servants over 25 years, found that habitually sleeping six hours or less per night at age 50 was associated with a 22 per cent higher risk of being diagnosed with dementia in later life; the figure was 37 per cent at age 60. The finding held up when sleep was measured directly with accelerometers rather than self-reported. In a UK-based, midlife-to-older cohort followed for a quarter of a century, this is about as clean a piece of evidence as the field provides.
There is a small fraction of the population — genetic estimates put it at under three per cent — who genuinely need less than six hours of sleep per night and feel fine on it. These ‘short sleepers’ carry specific genetic variants and have been studied in some detail. The vast majority of adults who claim to function fine on five hours are not short sleepers; they are sleep-deprived people who have lost the ability to perceive their own deficit.
The evidence for this last point is striking. In the landmark 2003 study by Van Dongen and colleagues, healthy volunteers were restricted to four, six or eight hours in bed for fourteen nights while cognitive performance was tested repeatedly. After fourteen nights at six hours, cognitive performance on tests of reaction time and sustained attention was equivalent to one night of total sleep deprivation. After fourteen nights at four hours, equivalent to two nights of total sleep deprivation. Crucially, the subjective sleepiness ratings of the six- and four-hour groups rose modestly and then plateaued: the subjects lost the ability to detect how impaired they actually were.
The under-slept lose the ability to detect their under-sleeping.
This is the single most important reason most adults systematically underestimate how much sleep they actually need. A related finding, from research at Australia’s Centre for Sleep Research, showed that after seventeen to nineteen hours of sustained wakefulness — a fairly ordinary late night — cognitive performance on tests of speed and accuracy is equivalent to, or worse than, performance at the UK drink-drive limit. After longer wakefulness, performance drops to the equivalent of the maximum legal alcohol dose in the study. Sleep deprivation and alcohol intoxication are not metaphorically comparable. They are physiologically comparable.
Sleep quality is the half of the equation that duration-focused conversations tend to ignore. Two adults can both report ‘about eight hours of sleep’ and one can be getting substantially more restorative sleep than the other. The difference sits in what sleep scientists call sleep efficiency — the percentage of time in bed actually spent asleep — and in how much of that sleep reaches the deeper, restorative stages.
The gap between time in bed and actual sleep is worth quantifying. A study of the CARDIA cohort that combined wrist actigraphy with self-reported sleep diaries found that adults overestimate their sleep duration by an average of 48 minutes — spending on average 6.8 hours of self-reported sleep but only 6 hours of measured sleep. The overestimate is larger at shorter sleep durations: adults averaging five hours of actual sleep typically report 1.2 hours more than they got; adults averaging seven hours report only 0.4 hours more. The practical implication is that an adult who ‘gets about six hours’ is probably getting closer to five. Most people are sleeping less than they think.
Several things reliably disrupt sleep quality without necessarily reducing duration. Alcohol is one of the largest. The most comprehensive review to date, published in Alcoholism: Clinical and Experimental Research in 2013 and covering 27 experimental studies, found that at all dosages, alcohol reduces the time to fall asleep (which is why it is widely used as an informal sleep aid) and consolidates the first half of sleep — but then increases fragmentation in the second half as it is metabolised. The effect on REM sleep is dose-dependent: low and moderate doses have unclear effects on first-half REM, but high doses markedly reduce it. What is consistent across all doses is the fragmentation and second-half disruption.
A member who has three glasses of wine in the evening and then sleeps for eight hours has not had eight hours of normal sleep; they have had something closer to four to five hours of deeper, first-half sleep, followed by three or four hours of lighter, more fragmented processing. The difference, cumulated across a month or a year, shows up in every metric of sleep-related health.
Caffeine consumed too late in the day is the second major quality disruptor. Caffeine’s half-life is around five to six hours in most adults, meaning an afternoon coffee is still actively present in the system at bedtime. A rigorous 2013 crossover study gave volunteers 400 milligrams of caffeine (roughly two large cups of coffee) at three different times — at bedtime, three hours before bed, and six hours before bed — and measured sleep with polysomnography and portable home monitors. The six-hours-before-bed condition produced disruption to sleep duration comparable to the at-bedtime condition. As the authors put it, the magnitude of sleep-time reduction supports the sleep-hygiene recommendation of no caffeine within six hours of bed. The practical rule of thumb: no caffeine after midday for most adults, and earlier still for those who are particularly sensitive.
Sleep quality also degrades when sleep is fragmented by noise, light, partner movement, children, or one’s own bladder. The cumulative effect of multiple brief awakenings — even ones the sleeper does not consciously remember — is a substantial reduction in restorative sleep. Experimental studies of sleep restriction in healthy young adults show that a week of five-hour nights reduces insulin sensitivity by 11 to 20 per cent, depending on measurement method, and produces measurable disruption to the hormones that regulate appetite — leptin drops, ghrelin rises. These are not abstract findings. They show up as hunger, carbohydrate craving, and subtle metabolic stress within days.
The immune system is similarly vulnerable. In a study at the University of California San Francisco, 164 healthy adults wore wrist actigraphs for a week, were then quarantined and given rhinovirus nasal drops, and monitored for cold symptoms. Adults sleeping less than five hours per night were four and a half times more likely to develop a cold than those sleeping more than seven. Adults sleeping between five and six hours were four times more likely. Between six and seven hours, the effect largely disappeared. Sleep, in this study, was a stronger predictor of getting a cold than any of age, income, stress, smoking status, or pre-existing antibody levels.
Almost every cell in the body keeps approximate time. There is a master clock in a small region of the brain called the suprachiasmatic nucleus, which sits just above the point where the optic nerves cross — and which receives direct input from the eyes about light exposure. The master clock then synchronises subordinate clocks in tissues throughout the body, telling each when to release particular hormones, when to repair, when to be most metabolically active. The whole system is called the circadian rhythm, and it is roughly tuned to a 24-hour day.
Light is, by a long way, the most powerful signal the circadian system responds to. Bright light in the morning, ideally within the first hour or two after waking, advances the clock and tells the body that the day has started. Dim, warm light in the evening allows the clock to release melatonin, the hormone that prepares the body for sleep. When this pattern is disrupted — bright artificial light in the evening, dim indoor light all day, screens before bed — the clock loses its bearings.
The evidence on evening screen use is particularly clear. A 2015 Harvard study compared reading on a light-emitting eReader for five nights versus reading a printed book for five nights, with a rigorous crossover design. The eReader condition suppressed evening melatonin by over 50 per cent, delayed the circadian phase (melatonin onset) by roughly one and a half hours, reduced REM sleep, and impaired next-morning alertness — an effect that persisted for hours after waking. The blue-wavelength light emitted by phones, tablets and laptops is what sleep scientists call the circadian zeitgeber problem: the evening light the eyes receive tells the body it is still daytime, and the clock shifts accordingly.
When the circadian system is misaligned with the actual sleep schedule — because of shift work, jet lag, late bedtimes, or evening light — the consequences extend beyond sleeping badly. Insulin sensitivity drops. Appetite regulation gets disrupted. Mood worsens. Cognitive performance degrades. The shift-work literature is the cleanest evidence of this: people who routinely work nights and sleep days have measurably worse health across almost every parameter studied, even when their total sleep duration is normal. The body cares not just whether you sleep, but whether you sleep at the right time.
Adults vary substantially in what the ‘right time’ is for them. A large epidemiological study using the Munich ChronoType Questionnaire, which has now been completed by more than 55,000 adults, found that sleep timing across the population follows a near-Gaussian distribution: most adults cluster around a middle range, with small tails at the extreme morning and extreme evening ends. Extreme early types wake up at roughly the time extreme late types fall asleep. Chronotype is both age-dependent (adolescents and young adults skew later, older adults earlier) and sex-dependent, and it has a substantial genetic component. For most adults this means the question is not ‘am I a lark or an owl’ but ‘what is my optimal sleep window, and does my current schedule respect it?’ A natural evening type forced to start work at 7am every morning is running at a persistent circadian disadvantage, and the health evidence suggests that disadvantage accumulates.
Until quite recently, sleep science treated sleep regularity — how consistently an adult goes to bed and wakes up at the same time each day — as a secondary concern next to duration. The evidence that has accumulated over the past five years has overturned that framing. For middle-aged and older adults, the regularity of sleep timing appears to be at least as important for long-term health outcomes as how many hours are slept.
The single most important study here is Windred and colleagues’ 2023 analysis of UK Biobank data, published in the journal Sleep. The study followed 60,977 adults who wore accelerometers for a week each, generating more than ten million hours of objective sleep-timing data. Over a median follow-up of 7.6 years, 1,859 participants died. The team calculated a Sleep Regularity Index — a measure of how similar an individual’s sleep-wake pattern is from one 24-hour period to the next — and tested it against all-cause mortality, cardiovascular mortality, and cancer mortality. Adults in the top four quintiles of regularity had between 20 and 48 per cent lower all-cause mortality than those in the least-regular bottom quintile. Critically, when the team controlled for total sleep duration, regularity remained a stronger independent predictor of mortality than duration itself.
A complementary finding comes from the US Multi-Ethnic Study of Atherosclerosis (MESA), where 1,992 adults underwent seven-day actigraphy and were then followed for a median of 4.9 years for incident cardiovascular events. Participants whose night-to-night variation in sleep duration exceeded two hours — or whose variation in sleep-onset timing exceeded an hour and a half — had more than double the risk of a cardiovascular event compared to those with the most regular patterns. Again, the effect held after adjusting for traditional cardiovascular risk factors and for average sleep duration.
The same story shows up in cardiometabolic markers. A study of 447 midlife day-shift workers found that social jetlag — the discrepancy between sleep-timing midpoints on workdays versus free days — was associated with lower HDL cholesterol, higher triglycerides, higher fasting insulin, greater insulin resistance, and greater adiposity, after adjusting for sleep duration, sleep quality, depression, and health behaviours. Lie in by three hours on a Sunday morning, repeat every weekend for decades, and the metabolic bill comes due.
Consistency of sleep timing matters at least as much as duration.
The practical implication is that bedtime consistency matters more than the dominant seven-to-nine-hours messaging would suggest. Going to bed and getting up within a fairly narrow window every day — including weekends — appears to be one of the most protective sleep behaviours available. The ‘Sunday night can’t sleep’ phenomenon that catches out so many adults after weekends of later nights is a small version of jet lag, and the chronic form, accumulated over decades, has measurable consequences.
Sleep hygiene is the slightly clinical name for the everyday levers an adult can pull to make good sleep more likely. Most are unsurprising in isolation. The cumulative effect of several of them, applied consistently, is substantial.
Temperature matters. The body needs to lose roughly one degree Celsius of core temperature in order to fall asleep, and a cooler bedroom — around 18 degrees Celsius is the typical recommendation — supports this. A room that is too warm is one of the most reliable causes of poor sleep, particularly in the second half of the night when body temperature naturally rises before waking. Bedding and nightwear that allow heat to dissipate help. This is especially worth getting right for perimenopausal and menopausal women, for whom hot flushes during the night are a common and disruptive complaint.
Light matters, again. Even modest light exposure during the night — a streetlight through curtains, a glowing standby light on a television, a phone screen checked at three in the morning — is enough to suppress melatonin and reduce sleep quality. A genuinely dark bedroom, with blackout curtains if needed, is among the cheapest and most effective single sleep interventions available. For members whose bedrooms cannot reliably be darkened — urban flats with thin curtains, travellers, shift workers who sleep during daylight hours — a well-fitting sleep mask is the practical alternative.
The light in the hour or two before bed matters almost as much. The blue and white wavelengths that dominate standard LED bulbs, overhead fluorescent lighting and most screens are the same wavelengths the circadian system is tuned to read as daytime. Switching to a warm-spectrum bulb in the bedroom and in the room where you spend the last hour of the evening — 2,700K or lower, ideally 2,200K or below — substantially reduces melatonin suppression. This is a one-off change that costs the price of a light bulb and works for as long as the bulb lasts. Members who make this single change often report falling asleep more easily within a week.
Screens before bed deserve particular attention because they combine three problems at once. The blue-wavelength light emitted by phones, tablets and laptops is particularly effective at suppressing melatonin, as the Harvard eReader study showed. The content — news, work email, social media — is typically activating rather than calming. And the simple act of being on a screen tends to delay the intended bedtime. Cumulatively, evening screen use is one of the largest correctable causes of bad sleep in modern adults. The first-best solution is to put screens away about an hour before bed. For members whose lives do not permit that — parents on call, people with evening work commitments, anyone whose partner wants to watch television — amber-tinted blue-light-blocking glasses worn in the hour before bed are a reasonable second-best. The evidence base for them is smaller than for simply reducing screen use, but what exists suggests they meaningfully reduce the melatonin suppression that blue-light exposure produces. They are not a replacement for good sleep hygiene. They are a mitigation for the unavoidable.
Alcohol and caffeine were covered under sleep quality and are worth repeating here as the two large correctable substances. Heavy meals close to bedtime are a smaller but real factor: the metabolic processing of a large meal interferes with sleep architecture, particularly with deep sleep in the early part of the night. A reasonable working rule is to finish eating two to three hours before bed. For adults who are sensitive to late eating, that window may need to extend to three or four.
Schedule consistency, which has its own sub-area above, belongs in the hygiene list too. Going to bed and waking up at roughly the same times every day — within an hour across the week — is one of the most powerful sleep interventions available, and one of the most-neglected by adults whose schedules vary substantially across weekdays and weekends.
The cumulative effect of getting these levers right is usually quite large. Adults who improve several of them simultaneously — cooler room, darker room, no screens for an hour before bed, no caffeine after midday, no alcohol within three hours of bed, consistent bedtime — typically see sleep quality improve markedly within a few weeks. Adults who try one at a time, with everything else unchanged, see smaller effects. The levers multiply each other rather than simply adding up.
Every Forever Well member receives a warm-spectrum bedside bulb (around 2,200K) as part of the standard membership. Silver and Gold members also receive a pair of amber-tinted blue-light-blocking glasses. Gold members additionally receive a sleep mask, for members whose bedrooms cannot be reliably darkened or who travel frequently. The hierarchy reflects the evidence: bulb first because ambient evening light is the common, universal problem; glasses next because evening screen use is widespread but not universal; mask last because the ability to achieve a dark bedroom is a problem for fewer members, but an acute one for those affected. None of these tools replaces the fundamentals — a cool bedroom, a consistent wake time, and the removal of caffeine and alcohol late in the day matter more than any of them. All three are shipped with the Month 3 sleep pack, alongside the pillar guide.
The six territories — architecture, duration, quality, circadian rhythm, regularity, and environment — are not independent. Disrupting one tends to disrupt the others. An adult with irregular sleep timing (going to bed at one in the morning on weekends, ten on weekdays) tends to have worse sleep quality (the circadian misalignment makes sleep less efficient) and may end up with shorter total duration too. An adult with a poor sleep environment (a hot, light bedroom) tends to lose deep sleep specifically, even if total time in bed is unchanged. Working on one lever helps. Working on several at once helps disproportionately more.
The structure of the rest of this pillar follows from this. Section 3 illustrates what poor and good sleep practice look like in the lives of two members. Section 4 sets out a practical framework for improvement, organised in the same three-tier structure as the Exercise pillar. Sections 5, 6 and 7 connect sleep to the other Forever Well pillars, lay out our editorial view, and point members toward further reading.
